Research On The Management Strategy Of The Underlying Data Buffer Zone In The Savant-Middleware-Based Internet Of Things

The popularization of the Internet of Things and the continual growth of the EPC system are rendering the Internet of Things gradually characteristic of real-time processing of huge amount of data in a complex network environment. However, the middleware of system is a weak computational platform with limited data processing capability. When the underlying data in the system cannot be effectively processed, the performance and availability of the whole system will be compromised. Therefore, it is imperative to examine and design the data processing in the middleware buffer.In this paper, the Savant-architecture-based technology of data acquisition, integration and screening, as well as buffer data processing is explored to tackle the above-mentioned problems. The results achieved are listed as follows:(1) By using the RED algorithm for reference, a data acquisition model based on dynamic strategy is proposed as a solution to such problems therein as the failure of subscribe-feedback mode to process "storms of data" and the complex management of enquiry-response system. The excellent performance of the algorithm lies in its relieving the network congestion in data acquisition yet retaining the advantages of the two data acquisition modes;(2) A data scheduling algorithm based on dynamic priority, with the capacity of real-time data processing, is proposed as a solution to such problems as the limited data processing capability of the middleware system, the contention of the readers for data resources arising from the rapidly growing data to be processed, the data of low priority "hungry" for processing in the strategy of the absolute priority and no dynamical adjustment of data priority allowed in the traditional hierarchical scheduling algorithm;(3)The redundant filter module, the feature filter module, and the abnormity filter module, which in combination constitute a flow-line controlled data filtering model that can be configured by some filter rules, are designed to eliminate the formerly inevitable redundant data and abnormal data in the middleware system and meet users’ occasional needs of data selection by specific criteria.Based on the above study, a simulation platform for system testing integrated with Savant system is set up, on which the aforementioned results are validated. The test shows that with the data processing module, data can be efficiently processed and scheduled into the buffer zone, and the labeled data can be effectively filtered, thus enhancing the performance and efficiency of the Savant system in terms of data processing.